Due to an increasing demand for collaborative robots, called "cobots", in industrial settings, this study aims to predict the chance of accidents occurring due to the introduction of cobots in the Korean manufacturing industry determined by a risk model applied Bayesian belief network. This will suggest effective risk mitigation measures. This study focuses on the types of safety monitored stop, as well as distance and speed control which have a higher collision chance compared to the types of power and force limiting which allow for injury-free contact and that of hand guiding which allows the cobot to move itself only by clear user's manipulation.

The factors that impact annual accident probability are built on the grounds of the analysis of occupational injuries and fatalities by industrial robots. These factors were then categorized into human, organizational, and technical errors. Each factor's probability was employed from the result of national statistics. If a probability was not available, notional probability was applied based on extensive literature reviews, and author's experiences over 10 years in the occupational safety and health fields due to it is scarce elsewhere.

The risk model is constructed with two decision nodes - the employer's and the policymaker's view - and twelve uncertainty nodes. The model showed that the estimated annual accident probability was the same as the average accident rate of the entire manufacturing industry of the Republic of Korea in 2018. This could be interpreted as "average-risky". Additionally, the influential factors were analyzed by a sensitivity analysis. By understanding which factors are highly influential, this study suggests three key measures to mitigate the risk by the introduction of cobots in the stages of design and manufacturing, installation, and usage. Researchers and OSH stakeholders may customize the model to assess the risk by the introduction of cobots.